Optical Microflaring on the Nearby Flare Star Binary UV Ceti

A&A 589, A48 (2016) Astronomy DOI: 10.1051/0004-6361/201628199 & c ESO 2016 Astrophysics

Optical microﬂaring on the nearby ﬂare star binary UV Ceti J. H. M. M. Schmitt1, G. Kanbach2, A. Rau2, and H. Steinle2

1 Hamburger Sternwarte, Gojenbergsweg 112, 21029 Hamburg, Germany e-mail: [email protected] 2 Max-Planck-Institut für extraterrestrische Physik, 85748 Garching bei München, Gießenbachstraße 1, Germany

Received 27 January 2016 / Accepted 11 March 2016

ABSTRACT

We present extremely high time resolution observations of the visual flare star binary UV Cet obtained with the Optical Pulsar Timing Analyzer (OPTIMA) at the 1.3 m telescope at Skinakas Observatory (SKO) in Crete, Greece. OPTIMA is a fiber-fed optical instrument that uses Single Photon Avalanche Diodes to measure the arrival times of individual optical photons. The time resolution of the observations presented here was 4 µs, allowing to resolve the typical millisecond variability time scales associated with stellar flares. We report the detection of very short impulsive bursts in the blue band with well resolved rise and decay time scales of about 2 s. The overall energetics put these flares at the lower end of the known flare distribution of UV Cet. Key words. stars: activity – stars: flare – stars: low-mass

1. Introduction conservative estimate of τth can be derived in the diffusion ap- proximation (which of course need not apply in the photosphere Solar and stellar flares are known to occur on vastly different and chromosphere) and one finds – apart from geometry factors time scales, ranging from a few seconds (Vilmer et al. 1994; CV (cf., discussion by Sánchez Almeida 2001)– tth ≈ 3 , where Schmitt et al. 1993) and possibly below (Beskin et al. 1982) to 16κRσT CV denotes the specific heat at constant volume, κR the Rosse- more than a week (Kürster & Schmitt 1996). The primary energy land mean opacity, σ the Boltzmann constant and T the tempera- release process of solar and stellar flares is – at least according ture. With typical parameters computed for a stellar photosphere to the prevailing paradigm – coronal, but a substantial amount of one finds cooling times on the order of a few seconds. If a sce- energy is (often) emitted at optical wavelengths. The response of nario of non-thermal heating and thermal cooling is assumed, the plasma, heated by particle beams and shock waves, depends one would thus expect cooling times on the order of seconds and sensitively on the mode of energy input and is in fact radiated heating time scales which in principle could be very short, i.e., over many regions of the electromagnetic spectrum. similar to the time scales of non-thermal solar X-ray or γ-ray Because of their low contrast white light flares are rather dif- bursts (Vilmer et al. 1994). ficult to observe on the Sun. However, on stars with sufficiently A thorough study of the photospheric heating (and cooling) low effective temperatures the photospheric response of a coro- processes during flares thus requires observations with very high nal energy release produces emissions that dominate the over- time resolution. Yet only few investigations of flare stars have all stellar output, even when confined to quite small portions of been carried out at extremely high time resolution at the level the stellar surface. Most researchers agree that in the flare pro- of, say, milliseconds or less. Since our OPTIMA instrument (for cess some particle acceleration does take place in the corona, a description see Sect.2) does provide this temporal resolution, causing the accelerated particles and/or plasma waves to move we decided to carry out some preliminary flare star observations along the magnetic field lines before finally reaching and heat- in the msec range and below. ing cooler atmospheric layers (Syratovskii & Shmeleva 1972; Of course, other flares star studies with high time resolution Brown 1973). The heated chromospheric and photospheric lay- have been carried out with various instruments. Robinson et al. ers are quite dense and start radiating essentially instantaneously, (1995) used the High Speed Photometer aboard the Hubble thus serving as a proxy indicator for non-thermal particles. On Space Telescope to study CN Leo with high time resolution somewhat longer time scales the heated chromospheric layers (0.01 s) for 2 h in the UV (around 2400 Å) and detected a num- “evaporate”, leading to a soft X-ray flare (Neupert 1978; Peres ber of events with “substantial variations sometimes occurring 2000). on time scales of less than 1 s”. Tovmassian et al.(1997) used If the optical emission were also non-thermal, the relevant smaller ground based telescopes to study the flare stars EV Lac time scales could be very short indeed. If, however, the emission and V 577 Mon with a time resolution of 0.1 s in the U and is thermal, as most researchers would assume, the relevant time B-band and argue that some of the reported events have durations scales are the typical propagation time of a shock τsh, driven by of less than 1 s. Similarly, Zhilyaev et al.(1998) report ground- the flare explosion into the photospheric layers, and the radiative based UBV observations again of EV Lac with a time resolution cooling time scale τth, on which a heated parcel of gas looses of down to 0.05 s, with some events having widths below 1 s. its energy by radiation. The shock propagation time can be es- The most systematic efforts we are aware of in that re- timated from the expression τsh ∼ H/vsh, with H denoting the spect are the studies undertaken at the 6 m Special Astrophys- scale height of the emission layer and vsh the shock speed. A ical Observatory (SAO) using the Multichannel Analysis of

Article published by EDP Sciences A48, page 1 of6 A&A 589, A48 (2016)

Nanosecond Intensity Alterations (MANIA); a detailed description of this hardware has been presented by Beskin et al.(1982). As with our OPTIMA instrument, MANIA recorded individual photon arrival times with a time resolution of 300 nsec. Most of the MANIA observations were taken with a Johnson U filter, some with a B filter, however, due to data storage limitations at the time only the strongest flares could be recorded for subsequent analysis. Needless to say, data storage is not an issue any more. The MANIA team published data on a number of nearby flare stars (Beskin et al. 1988), in particular on UV Cet, here we present OPTIMA data on the nearby (d = 2.62 pc) flare star binary UV Cet (=Gl 65 A+B). According to the SIM- BAD data base, both components are spectral type dM5.5e, the brighter component has a visual magnitude of 12.57 mag with a B − V color of 1.85 mag, while the B component with a visual magnitude of 12.7 mag is only a little fainter. Both components are dMe stars, both are known to be active, and both are known to produce flares. Both binary components have also been 20" detected at X-ray (Audard et al. 2003) and radio wavelengths (Jackson et al. 1989), with the B component usually being the more active one. According to Geyer et al.(1988) the two com- Fig. 1. Image of the UV Cet system taken with the OPTIMA field view- ponents orbit each other in a rather eccentric orbit with a period ing CCD camera. The positions of the optical fibers are indicated rela- of 25.52 yr; at the time of our observations the two stars were tive to the location of the target. separated by a little under 2 arcsec at a position angle of 50◦.

100 2. Observations and data reduction 80 The observations reported in this paper were taken with the Op- 60 tical Pulsar Timing Analyzer (OPTIMA) mounted at the 1.3 m 40 Johnson-B OPTIMA telescope at Skinakas Observatory (SKO) in Crete, Greece in 20 2008. OPTIMA is a transportable visiting instrument that was (%) Transmission 0 350 400 450 500 550 operated at SKO for several weeks each year between 2006 and 30 2012. It records the arrival times of individual optical photons 25 using eight single photon avalanche diodes (SPADs). Each diode 20 is fed with by an optical fiber mounted in the telescopes optical 15 plane on a slanted mirror. A central (“target”, cf., Fig.1) fiber is 10 surrounded hexagonally by six outer fibers, while an eighth fiber (%) Efficiency 5 observes the sky background a few arcmin away from the target. 0 The observations discussed here have been performed with 350 400 450 500 550 Wavelength (nm) a set of 300 µm diameter fibers corresponding to 6 arcsec on the sky; thus it is clear that the two components of UV Cet cannot Fig. 2. Top: comparison of the transmission of blue filter used within be resolved by OPTIMA. The data acquisition system used in OPTIMA and the standard Johnson B-band filter. Bottom: overall effi- 2008 recorded all photons from the eight SPADs together with ciency of OPTIMA with the blue filter taking into account the quantum a time stamp accurate to 4 µs and synchronized with a GPS sig- efficiency of the single photon avalanche diodes. nal in absolute time. At this time resolution dead time effects become important at count rates above ≈105 ct/s. The data were stored in a buffer, which, once full, was written to a hard drive. During this writing process no further data were recorded lead- the observations this sequence is reversed and another set of sky ing to an additional dead time of several seconds approximately background and SPAD dark current measurements is taken. every hour. In addition, to the photons recorded by the SPADs photons arriving outside the fibers are reflected on the slanted A blue filter can be inserted into the light path of the fibers; mirror and imaged with a CCD camera. Here, series of expo- the filter is transmissive in a relatively narrow wavelength range sures with integration times of 10 s are obtained to monitor between 4000 Å and 4600 Å, while the quantum efficiency of for variations in atmospheric transmission and seeing during an the SPADs peaks at ≈7000 Å with very little transmission below OPTIMA observation. 4000 Å. For the later analysis it is important to realize that this A typical OPTIMA observing sequence of a source starts filter does not reproduce any of the standard color systems. How- with dark current measurements of the SPADs for about three ever, its effective transmission curve resembles that of the stan- minutes (cf., Fig.3). Then the telescope is slewed towards a dard B filter in the Johnson UBV system (see Fig.2). The above source-free position in the vicinity of the target for sky back- summarizes the instrument as it was used for the UV Cet ob- ground measurements, again typically for a about three min- servations in 2008. More details can be found in Kanbach et al. utes. Afterwards the telescope is repositioned onto the target (2008). Since then, OPTIMA has continuously evolved and its and the actual science data taking commences. At the end of performance has been upgraded.

A48, page 2 of6 J. H. M. M. Schmitt et al.: Optical microﬂaring on the nearby ﬂare star binary UV Ceti

Fig. 3. OPTIMA light curve of UV Cet (bottom panel) and a nearby sky background position (top panel) taken on October 7th 2008. The flares Fig. 4. Comparison of observed distribution of counts per millisecond of interest are indicated by the dotted lines; see text for more details. (solid histogram) for the second hour of OPTIMA data of UV Cet with a Poisson distribution with the observed mean (solid curve). 3. Results 3.1. OPTIMA observations of UV Cet of 1 ms; the corresponding count histogram (constructed from 3.6 × 106 samples) together with a Poisson distribution with the During our 2008 OPTIMA observing campaign at SKO we ob- same expectation value is shown in Fig.4, which demonstrates served the UV Cet binary system on October 6th and 7th for that the the Poisson distribution does provide a good description about 3 h each. Useful data with stable pointing were obtained of the data. We do, however, mention in passing that an inspec- for about 5000 s in the first night and 10 000 s in the second tion of the moments of the observed count distribution shows night. that its variance, skewness and kurtosis are somewhat larger than Given the binary separation of below 2 arcsec and the expected for a pure Poissonian distribution with the same mean, OPTIMA fiber size of 6 arcsec, it is clear that the two binary the values of kurtosis being about 8% larger. Obviously, the ob- components cannot be separated with the available instrumental served distribution has more outliers than a pure Poissonian case. setup as demonstrated in Fig.1, where we show the hexagonal As we are interested in the source intrinsic variability on time OPTIMA fibers as appearing on the sky. In order to give an im- scales of a few milliseconds all other potential contributions to pression of the quality of our OPTIMA data on UV Cet, we show the variability as well as to the uncertainties must be investi- in Fig.3 the raw data (binned to a time resolution of 1 s) obtained gated. The scintillation phenomenon is well known to produce in the central channel during the second night (bottom panel) as variability on these time scales of interest (Dravins et al. 1997). well as a background channel close to the target (top panel); note Scintillation, i.e., the “twinkling of stars”, is caused by turbu- that the zero point of the time refers to UTC 21:02:34, Oct. 7, lence in the high atmosphere at altitudes of ten or more kilome- 2008. One recognizes very clearly the dark measurements at the ters. Inhomogeneities in temperature and density lead to inho- beginning and at the end of the UV Cet run and, further, the sky mogeneities in the refractive index, which cause random phase background continuously decreased during the first 20 min of changes in a transmitted wave field of a celestial source. As a the observations (the Moon was setting at the time), however, its consequence, the wave field is distorted and the illumination on level amounts to less than 5% of the overall source signal. the ground will be inhomogeneous, too. A telescope with some The data drops at approximately 3600 s and 7200 s are given aperture will capture variable amounts of light. With in- caused by reading out of the full buffer when no new data can creasing aperture one integrates over a larger number of atmo- be stored as described above. The data drop near T ≈ 8800 s is spheric turbulence elements and therefore effectively depresses caused by a loss of pointing and subsequent reacquisition of the scintillation effects. According to theory the recorded scintilla- target. tion intensity is expected to follow a log-normal distribution. An inspection of the data obtained in the other channels This distribution is sampled in the observation as winds in the shows that these channels are dominated by background, how- upper atmosphere transport the turbulent elements across the ever some leakage does occur as apparent from the data drop telescope aperture. On time scales sufficiently short the atmo- during the auto guider loss. However, that leakage is smaller than sphere is ”frozen-in”, since all light has traversed the same reali- the additional background in these channels and we therefore re- sation of turbulent elements, and different samples are not statis- frain from adding channels. Since we are performing only rela- tically independent, leading to an autocorrelation of the recorded tive photometry some losses through leakage are immaterial. In stellar signals. Fig.3 two events are marked as flares at times of about 6450 s We therefore checked the autocorrelation structure of our and 9370 s, which at frist are not very conspicuous but are actu- UV Cet data. In order to construct the autocorrelation function ally significant short duration flares on UV Cet as we will show (ACF), we considered light curve chunks of 10 s each. For each below. of these intervals the ACF was computed for time lags of multi- ples of 1 ms and a mean ACF was computed. In Fig.5 we plot 3.2. Overall variability on short time scales the resulting mean ACF which is seen to indeed increase towards small time lags as expected from scintillation, however, the cor- We first inspect our data for the properties of the event statistics. relation effects are still relatively small. We carried out a similar We specifically consider one hour of data without any instru- analysis on the background channels and found no significant mental interruptions and consider a data sampled with a cadence autocorrelation values on the relevant time scales.

A48, page 3 of6 A&A 589, A48 (2016)

Fig. 6. OPTIMA light curve of flare 1 at a time resolution of 10 ms (top Fig. 5. Mean autocorrelation function ACF vs. lag (in ms) for the second panel) together with an empirical fit (see text for details) and median hour of OPTIMA data of UV Cet. The sharp rise towards small lag times averaged light curve (bottom panel). indicates the presence of scintillation effects, the slow decay indicates unaccounted fluctuations. Table 1. Measured quantities and derived physical parameters for the two flares observed; see text for details. In summary, we can state that on time scales of 10 ms our OPTIMA light curves of UV Cet show some small degree (about Parameter Flare 1 Flare 2 5%) of autocorrelation, which we attribute to scintillation effects. In time bins of 10 ms we typically find about 40 source count Peak count rate (s−1) 4000 2000 (cf., Fig.3.) out of which 1 −2 are unrelated to the source. The Rise time (s) ≈2 ≈1.5 observed count distributions resemble Poisson distributions with Fast decay time (s) ≈2 ≈1.5 some extra noise. The background channels show no signatures Slow decay time (s) ≈13 n.a. of auto- or crosscorrelation effects as expected. Counts in rise ≈4000 ≈1600 ≈ ≈ In the following we concentrate on the source channel and Counts in fast decay 3900 3800 ≈ specifically on the two flares observed in the night of Oct. 7th Counts in slow decay 600 n.a. 2008 (see Fig.3). The brighter flare (henceforth Flare 1) oc- Quiescent rate (cts/s) 3700 3700 curred at 9370 s while the fainter peak (Flare 2) happened at Peak luminosity (erg/s) 1.6 × 1028 7.9 × 1027 6450 s. No flares were detected in the observations from Oct. 6th Energy (rise, erg) 1.6 × 1028 6.3 × 1027 2008. Energy (decay fast) 1.5 × 1028 1.5 × 1028 Energy (decay slow) 2.4 × 1027 n.a. Energy (total) 3.3 × 1028 2.1 × 1028 3.3. Flare 1 Area (cm2) 2 × 1016 1.4 × 1016 In Fig.6 (lower panel) we show an excerpt of our UV Cet light curve at a time resolution of 10 ms. In order to take out trends and short term oscillations we take chunks of typically 50−100 s of data prior to and after the flare and fit a polynomial of order The flare light curve exhibits multiple flux drops with a of up to ten to these data. We then use the fitted polynomial to quasi-regular spacing of about four seconds. This pattern is only rectify the OPTIMA data. In order to reduce the noise in the vi- detected during the flare and not in the non-flaring UV Cet light sualisation of the light curves we also apply a sliding median av- curve data. No such behavior has been seen in any other source eraging over 250 ms and show the thus obtained smoothed light observed with OPTIMA and no potential instrumental features curve in the bottom panel of Fig.6, where the typical signature appearing with that period are known. Thus, the probability is of a flare becomes very much apparent. Additionally, we show high that the observed quasi periodicity is source intrinsic. Ex- an adhoc model fit to the data consisting of a linear rise phase, perimental verification of such oscillations in other flares is cer- followed by two exponential decay phases. This is meant only as tainly highly desirable. an empirical description of the data to specify peak flux, fluence and decay time scales of this event; in Table1 we provide these basic parameters of the flare. 3.4. Flare 2 Flare 1 looks like a textbook example of a stellar flare rising steeply to reach its peak within two seconds. The peak lasts only Flare 2 was analyzed with the same procedures as for flare 1. The a very short time, at most 1−2 s, afterwards the flux enhancement resulting light curve (shown in Fig.7) shows that the morphol- decreases initially very fast with an estimated decay time scale ogy of flare 2 is very similar to that of flare 1. However, flare 2 of about two seconds. A definite break is apparent in the light is less bright and thus the slow decay section of the light curve is curve and a much slower decay on a time scale of about 15 s not readily apparent in the unbinned data. The median averaged follows. Since it is difficult to determine the zero level with very light curve does suggest an enhanced level lasting for about eight high precision, this latter time scale and hence the total flux in seconds. The resulting fit parameters for flare 2 are also included this slow decay part are therefore less well determined. in Table1.

A48, page 4 of6 J. H. M. M. Schmitt et al.: Optical microﬂaring on the nearby ﬂare star binary UV Ceti

luminosity in the blue band (in terms of erg/s) and the total emitted B band energy. As the temperature of the flare plasma is bound to change with respect to quiescent conditions, our energy conversion can naturally only be approximately correct. We note that while under quiescent conditions the blue luminosity accounts for only a small fraction of the overall energy losses, the flaring regions have considerably higher luminosities. Thus, the blue band should actually capture a substantial fraction of the overall optical energy losses. Needless to say, no statements about energy losses in other than the observed band can be made, thus our estimates should be considered as lower limits. Taking the observed values as given in Table1, we can derive the energetics of the observed flare events. Here we concentrate on the rise and rapid decay phase, as the slow decay phase is dif- Fig. 7. OPTIMA light curve of flare 2 at a time resolution of 10 ms (top ficult to discriminate for the weaker flare and the time scale and panel) together with an empirical fit (see text for details) and median hence the total fluence are poorly constrained. As data were only averaged light curve (bottom panel). obtained with a single filter the temperature of the flaring plasma cannot be determined. However, from other flare observations we know that typical temperatures are on the order of 10 000 K. 3.5. Comparison to MANIA observations With such temperatures any bolometric corrections ought to be The flare patrol time of our OPTIMA UV Cet observations small. Therefore the derived numbers should be representative was considerably shorter than the >35 h patrol time reported by for the whole energy emitted in the optical regime. Naturally Beskin et al.(1988) for MANIA. In addition, the large majority we cannot make any statements on energy losses in other wave- of the MANIA observations were obtained with the U-band filter bands, which are likely to occur and thus our numbers should be and only a few were taken in the B-band, similar to our passband. considered as lower limits. From the determined peak luminosity we then derive a char- Unsurprisingly, Beskin et al.(1988) thus reported 118 U-band, 16 2 but only 4 B-band flares. As no detailed analysis of all ob- acteristic area of 1–2 × 10 cm as the emitting surface on the served MANIA flares is available, we can only state that some star, assuming a canonical effective temperature of 10 000 K. of their flares very much resemble the two events observed with While clearly these estimates should be considered only as order OPTIMA. of magnitude estimates that depend sensitively on temperature, it is clear that the characteristic size of the flaring regions is quite small with an extent of less than one thousand kilometers. 3.6. Energetics

In the following section we provide a rough estimate of the ener- 4. Discussion and conclusions getics of the two flares observed with OPTIMA. In this context we note again that the blue filter used for our OPTIMA observa- The peak luminosities of the flares recorded in our OPTIMA tions is not a standard filter, however, we assume for simplicity observations reach about 50−100% of the star’s quiescent B- that the filter response is that of a Johnson B filter. This assump- band luminosity. Their total energy output equals that emitted tion should not be crucial since we cannot determine the tem- by the star in a few seconds in the B-band, yet they clearly make perature of the flare plasma from our single band observations only very small contributions to the overall energy output of the anyway, and therefore the overall energetics can be determined star, given its bolometric luminosity of 6 × 1030 erg s−1. The only very approximately. For the purposes of our discussion we recorded fluctuations make a reliable identification of shorter assume a bolometric magnitude of both components of UV Cet lasting or weaker flares difficult. We are, however, confident that of 11.77 (Greenstein 1989), which corresponds to a bolometric in the total OPTIMA patrol time on UV Cet (a total of less than 30 −1 luminosity at UV Cet’s distance of Lbol ≈ 6.8 × 10 erg s . five hours) no stronger flares than the above reported flares 1 Furthermore we assume that an apparent magnitude of mb = 0 and 2 occurred. corresponds to an energy flux outside the Earth’s atmosphere of With our overall, rather short patrol time on UV Cet we can 6.49 × 10−6 erg cm−2 s−1. Using these numbers as well as an ap- state that the average B-band energy input is on the order of parent B magnitude of 13.86 we compute an energy flux (in the 3−4 × 1024 erg s−1 with significant uncertainties. This number blue band) of 1.85×10−11 erg cm−2 s−1 and a total blue luminos- is two orders of magnitude smaller than the U-band value as ity of 1.5 × 1028 erg s−1 for the UV Cet binary system. quoted by Doyle & Butler(1985). To determine the quiescent source photon flux, we determine Lacy et al.(1976) studied the overall energetics of flares ob- the median dark SPAD count rate from the exposure with closed served on different stars in different energy bands and found shutter, estimate the sky background from the fibers surrounding the relation EU = (1.2 ± 0.08)EB over several orders of magni- the target fiber and correct for individual fiber efficiencies. As the tude. Using this relation suggests that our B-band flare “survey” SPAD counts in the central fiber are dominated by source counts, is highly incomplete and misses substantial numbers of small these corrections are quite small. Associating the observed qui- flares. Observations in the U-band should reduce photospheric escent photon flux observed before and after the flares with an “background” by a factor of about 3 or more and correspond- energy flux of 1.85 × 1011 erg cm−2 s−1 results in a luminosity ingly increase our sensitivity to detect even smaller flares. The to count rate conversion factor of 3.95 × 1024 erg/count for our energetics derived in the B-band puts the flares – not surpris- OPTIMA UV Cet data. Assuming then that this conversion fac- ingly – at the low end of the observed flare energetics distribution tor is the same also for the flare data, we can assess the peak derived by Lacy et al.(1976) for UV Cet.

A48, page 5 of6 A&A 589, A48 (2016)

The rise times of the observed flares were on the order of findings of the MANIAC team that the typical time scale for stel- two seconds. Shvartsman et al.(1988) report the rise times of lar flares (in the optical) can be in the range 0.1−1 s, although 10 flares observed on UV Cet, two of those had rise times com- most flares have – just like “our” flares on UV Cet – overall du- parable to or even shorter than the rise times reported for flares 1 rations in excess of 1 s. Therefore time resolutions in the range and 2. Thus, the observed rise times are short, yet not unusually 1−10 ms are sufficient for a full description of the relevant phe- so. It is not clear whether the observed rise times should be color nomena. We finally note high spectral resolution observations of dependent. High spectral resolution observations of a giant flare flare events with possibly sub-second cadence might provide a on CN Leo suggest a markedly different behavior of the flaring science case for very large telescopes. We expect the occurrence plasma during the impulse and gradual phase (Fuhrmeister et al. of spectral changes both in the form of temperature changes as 2008). However, this giant flare on CN Leo may not be compa- well as in the appearance and disappearance of emission lines on rable to the by comparison rather tiny flares discussed here. short time scales and a theoretical modelling of such processes, There are various reports about apparent flux oscillations requiring a combined treatment of hydrodynamics and radiative for active late type dwarf stars. A still prominent example is transfer, is now within reach of numerical astrophysics. a long duration flare on the Hyades star H II 2411 observed by Rodono(1974), showing coherent oscillations with a period of 13 s throughout its eruption. Further, Zhilyaev et al.(2000) Acknowledgements. This research has made use of the SIMBAD database, oper- report the results on multi-site, multi-band observations of the ated at CDS, Strasbourg, France. Skinakas Observatory is a collaborative project flare star EV Lac, finding “coherent” oscillations in three out of the University of Crete (UoC) and the Foundation for Research and Technol- ogy Hellas (FORTH), and the Max-Planck-Institute for Extraterrestrial Physics. of 19 recorded flares with periods in the range 13 s and 26 s This work was in part supported under the FP7 Opticon European Network for on rather long duration flares lasting more than 100 s. Unfor- High Time Resolution Astrophysics (HTRA) project. We want acknowledge the tunately these authors do not discuss flares with shorter dura- contributions of Natalia Lewandowska (UHH) and Natalia Primak (IfA), who tions, although the time resolution of their data (0.1 s) would participated in the observing campaigns yielding the data analyzed in this paper. have allowed them to do so. At any rate, if we accept the observed oscillations in flare 1 as real and associate the observed period τ with the size and Alfven speed through τ = L/vA, we References find – using L ∼ 108 cm, B ∼ 103 G and τ ∼ 4 s densities n ∼ 1014 cm−3, which fit to the expected chromospheric origin Audard, M., Güdel, M., & Skinner, S. L. 2003, ApJ, 589, 983 Beskin, G. M., Neizvestnyi, S. I., Pimonov, A. A., Plakhotnichenko, V. L., & of the flare radiation. Shvartsman, V. F. 1982, IAU Colloq. 67: Instrumentation for Astronomy with It is instructive to compare the energetics derived for the Large Optical Telescopes, 92, 181 optical flares to the X-ray data for UV Cet. UV Cet’s quies- Beskin, G. M., Chekh, S. A., Gershberg, R. E., et al. 1988, Sov. Astron. Lett., 14, cent flux in the soft X-ray band is typically on the order of 2– 65 × 27 −1 Betta, B., Peres, G., Reale, F., & Serio, S. 1997, A&AS, 122, 585 3 10 erg s , and for comparison, the potentially weakest Brown, J. C. 1973, Sol. Phys., 28, 151 X-ray flare from UV Cet reported to date (Schmitt et al. 1993) Cargill, P. J., & Klimchuk, J. A. 2004, ApJ, 605, 911 had a total emitted energy of 1.5 × 1029 erg, i.e. on the same or- Dravins, D., Lindegren, L., Mezey, E., & Young, A. T. 1997, PASP, 109, 173 der of the total energy loss of the optical flares reported in this Doyle, J. G., & Butler, C. J. 1985, Nature, 313, 378 paper. Fuhrmeister, B., Liefke, C., Schmitt, J. H. M. M., & Reiners, A. 2008, A&A, 487, 293 Heating by so-called “microflares” and “nanoflares” Geyer, D. W., Harrington, R. S., & Worley, C. E. 1988, AJ, 95, 1841 (Cargill & Klimchuk 2004; Klimchuk 2006) is a popular hypoth- Greenstein, J. L. 1989, PASP, 101, 787 esis to explain the heating of the corona of the Sun and that of Jackson, P. D., Kundu, M. R., & White, S. M. 1989, A&A, 210, 284 other stars. Their nanoflare model assumes that any coronal loop Kanbach, G., Stefanescu, A., Duscha, S., et al. 2008, Astrophys. Space Sci. Libr., 351, 153 is composed of many unresolved strands of magnetic flux and Klimchuk, J. A. 2006, Sol. Phys., 234, 41 that these strands are “heated impulsively by a small burst of en- Kürster, M., & Schmitt, J. H. M. M. 1996, A&A, 311, 211 ergy, which, for convenience, are called a nanoflare, although the Lacy, C. H., Moffett, T. J., & Evans, D. S. 1976, ApJS, 30, 85 range of energies is completely arbitrary”. However, the energies Neupert, W. M. 1968, ApJ, 153, L59 of the flare events observed on UV Cet are still orders of mag- Moffett, T. J. 1972, Nat. Phys. Sci., 240, 41 Parker, E. N. 1988, ApJ, 330, 474 nitude above the nanoflare energetics discussed in the context of Peres, G. 2000, Sol. Phys., 193, 33 heating the solar corona. At any rate, it would be interesting to Robinson, R. D., Carpenter, K. G., Percival, J. W., & Bookbinder, J. A. 1995, explore a possible connection of optical microflaring to analo- ApJ, 451, 795 gous X-ray activity in a more systematic way, and we therefore Rodono, M. 1974, A&A, 32, 337 Sánchez Almeida, J. 2001, ApJ, 556, 928 encourage co-temporaneous X-ray and optical observations with Schmitt, J. H. M. M., Haisch, B. M., & Barwig, H. 1993, ApJ, 419, L81 a high time resolution. Shvartsman, V. F., Beskin, G. M., Gershberg, R. E., Plakhotnichenko, V. L., & In summary, OPTIMA is well suited for the study of rapid Pustilnik, L. A. 1988, Sov. Astron. Lett., 14, 97 variability on late-type stars. Clearly, the usage of a bluer filter Syratovskii, S. I., & Shmeleva, O. P. 1972, Sov. Astron., 16, 273 (e.g., U-band) together with SPADs with enhanced quantum effi- Tovmassian, H. M., Recillas, E., Cardona, O., & Zalinian, V. P. 1997, Rev. Mex. Astron. Astrofis., 33, 107 ciency at lower wavelengths would greatly enhance our ability to Vilmer, N., Trottet, G., Barat, C., et al. 1994, Space Sci. Rev., 68, 233 detect events of even lower luminosities at shorter wavelengths. Zhilyaev, B. E., Verlyuk, I. A., Romanyuk, Y. O., et al. 1998, A&A, 334, 931 Our (currently) sparse observational material confirms previous Zhilyaev, B. E., Romanyuk, Y. O., Verlyuk, I. A., et al. 2000, A&A, 364, 641

A48, page 6 of6